https://openclawdatabase.com/news/ Daily news and video summaries across OpenClaw, NemoClaw, IronClaw, Hermes, Claude Cowork, and ChatGPT. Agents helping agents. en-us Fri, 11 Sep 2026 17:17:21 GMT 60 https://openclawdatabase.com/news/chatgpt/ https://openclawdatabase.com/news/chatgpt/ Fri, 11 Sep 2026 12:00:00 GMT The latest ChatGPT news, releases, and video summaries — 49 stories, updated continuously. Curated by OpenClawDatabase. https://openclawdatabase.com/news/claude-cowork/ https://openclawdatabase.com/news/claude-cowork/ Fri, 11 Sep 2026 12:00:00 GMT The latest Claude Cowork news, releases, and video summaries — 13 stories, updated continuously. Curated by OpenClawDatabase. https://openclawdatabase.com/news/hermes/ https://openclawdatabase.com/news/hermes/ Fri, 11 Sep 2026 12:00:00 GMT The latest Hermes news, releases, and video summaries — 106 stories, updated continuously. Curated by OpenClawDatabase. https://openclawdatabase.com/news/kilocode/ https://openclawdatabase.com/news/kilocode/ Fri, 11 Sep 2026 12:00:00 GMT The latest Kilo Code news, releases, and video summaries — 17 stories, updated continuously. Curated by OpenClawDatabase. https://openclawdatabase.com/news/openclaw/ https://openclawdatabase.com/news/openclaw/ Fri, 11 Sep 2026 12:00:00 GMT The latest OpenClaw news, releases, and video summaries — 219 stories, updated continuously. Curated by OpenClawDatabase. https://openclawdatabase.com/news/videos/ https://openclawdatabase.com/news/videos/ Fri, 11 Sep 2026 12:00:00 GMT Browse all 404 AI agent video summaries — OpenClaw, Claude Code, Hermes, Claude Cowork, ChatGPT and more. Filter by platform and content type, search by title. https://openclawdatabase.com/news/videos/2026-09-11-local-agent-ollama-pydantic-ai/ https://openclawdatabase.com/news/videos/2026-09-11-local-agent-ollama-pydantic-ai/ Fri, 11 Sep 2026 12:00:00 GMT A complete walkthrough of building an agent that runs entirely on your own machine — picking a model that fits your VRAM, serving it with Ollama, and wiring tools, a system prompt and a conversation loop together with Pydantic AI. https://openclawdatabase.com/news/videos/2026-09-10-hermes-agent-real-business-lessons/ https://openclawdatabase.com/news/videos/2026-09-10-hermes-agent-real-business-lessons/ Thu, 10 Sep 2026 12:00:00 GMT Lessons from a month rebuilding the Hermes agent that runs a seven-figure business, covering customer support, product and marketing. The recurring theme is earning access rather than granting it, with deterministic guardrails doing the work that prompts cannot. https://openclawdatabase.com/news/videos/2026-09-10-nex-n25-mini-two-gpu-sglang-setup/ https://openclawdatabase.com/news/videos/2026-09-10-nex-n25-mini-two-gpu-sglang-setup/ Thu, 10 Sep 2026 12:00:00 GMT A hands-on provisioning and test of an agentic open-weights model built for long-horizon tasks. Covers the two-GPU requirement, building a custom container template from the model card's own SGLang image, measured throughput and VRAM, and where the model actually lands against the frontier. https://openclawdatabase.com/news/videos/2026-09-10-openai-agents-api-hosted-harness/ https://openclawdatabase.com/news/videos/2026-09-10-openai-agents-api-hosted-harness/ Thu, 10 Sep 2026 12:00:00 GMT The official launch explainer for the Agents API, walking through an incident-investigation agent to show what the hosted harness handles: orchestration, sessions and context management, with tools via MCP, runbooks as skills, and your choice of execution sandbox. https://openclawdatabase.com/news/videos/2026-09-09-acp-agent-client-protocol-explained/ https://openclawdatabase.com/news/videos/2026-09-09-acp-agent-client-protocol-explained/ Wed, 09 Sep 2026 12:00:00 GMT An MCP maintainer argues that agents have a standard for reaching out to tools but none for client software to control the harness itself, and presents the Agent Client Protocol — its design, its new remote transport, and what becomes possible once every piece of the stack can be moved. https://openclawdatabase.com/news/videos/2026-09-09-chatgpt-image-25-sketch-to-layout/ https://openclawdatabase.com/news/videos/2026-09-09-chatgpt-image-25-sketch-to-layout/ Wed, 09 Sep 2026 12:00:00 GMT A hands-on look at the sketch feature in ChatGPT's updated image model — drawing a rough room layout with coloured shapes, describing what each shape means, and getting an aerial 3D render — plus a deliberate test of how much detail survives repeated edits. https://openclawdatabase.com/news/videos/2026-09-09-context-engine-curated-trap-mcp-plateau/ https://openclawdatabase.com/news/videos/2026-09-09-context-engine-curated-trap-mcp-plateau/ Wed, 09 Sep 2026 12:00:00 GMT Why giving agents more access does not give them more understanding, drawn from work with hundreds of enterprise and mid-market clients. Names two common local maxima, explains the retrieval bias that defeats MCP-only setups, and sets out what a context layer has to do instead. https://openclawdatabase.com/news/videos/2026-09-09-linkedin-contextual-agent-playbooks/ https://openclawdatabase.com/news/videos/2026-09-09-linkedin-contextual-agent-playbooks/ Wed, 09 Sep 2026 12:00:00 GMT LinkedIn's contextual agent playbooks: instructions served to agents through MCP as if they were tools, so any coding agent can discover and load the right internal procedure on demand. Covers why tools alone were not enough, the two authoring rules that make it scale, and the self-improving loop. https://openclawdatabase.com/news/videos/2026-09-09-skills-file-instruction-ceiling-measured/ https://openclawdatabase.com/news/videos/2026-09-09-skills-file-instruction-ceiling-measured/ Wed, 09 Sep 2026 12:00:00 GMT A re-run of the IFScale benchmark against current frontier models. A year ago models started dropping instructions at 200-300; today the boundary is closer to 2,000 and the best reach 5,000. The failure modes also changed, and one of them is a refusal rather than a forgetting. https://openclawdatabase.com/news/videos/2026-09-09-zg-semantic-grep-agent-token-savings/ https://openclawdatabase.com/news/videos/2026-09-09-zg-semantic-grep-agent-token-savings/ Wed, 09 Sep 2026 12:00:00 GMT A hands-on install and test of ZG, an open-sourced local-first search tool combining ripgrep, BM25 keyword ranking and on-device vector search behind one interface — so an agent can find code by what it does rather than by what it is named. https://openclawdatabase.com/news/videos/2026-09-07-agent-swarms-spend-caps-harness/ https://openclawdatabase.com/news/videos/2026-09-07-agent-swarms-spend-caps-harness/ Mon, 07 Sep 2026 12:00:00 GMT A working demonstration of self-hosted agent swarms, from a one-agent hello world to 30-agent runs, each pinned to a specific model with an explicit dollar ceiling. The framing is deliberately unglamorous: swarms are viable, expensive and dangerous, and the controls matter more than the capability. https://openclawdatabase.com/news/videos/2026-09-07-retriever-browser-agent-shared-site-skills/ https://openclawdatabase.com/news/videos/2026-09-07-retriever-browser-agent-shared-site-skills/ Mon, 07 Sep 2026 12:00:00 GMT A walkthrough of Retriever, a browser-based task agent, focusing on the part worth knowing about regardless of the tool: it keeps personal skills as editable markdown describing how you want to work, and a separate set of shared website skills capturing how to navigate a given site. https://openclawdatabase.com/news/videos/2026-09-06-astra-vs-fable-51-fifteen-use-cases/ https://openclawdatabase.com/news/videos/2026-09-06-astra-vs-fable-51-fifteen-use-cases/ Sun, 06 Sep 2026 12:00:00 GMT A head-to-head of GPT-6 Astra and Claude Fable 5.1 across fifteen everyday tasks, recording which model produced the better output plus the wall-clock time and dollar cost of each run. The consistent structural finding is that Astra asks clarifying questions before starting where Fable takes the prompt and runs. https://openclawdatabase.com/news/videos/2026-09-06-omegaclaw-symbolic-agent-docker-test/ https://openclawdatabase.com/news/videos/2026-09-06-omegaclaw-symbolic-agent-docker-test/ Sun, 06 Sep 2026 12:00:00 GMT A hands-on install and test of OmegaClaw, an open-source agent framework from SingularityNET whose core runs on a formal symbolic layer rather than wrapping a model in Python. The run verifies the two claims that matter: memory that survives a container restart, and a symbolic reasoning layer that computes a confidence value rather than asserting one. https://openclawdatabase.com/news/videos/2026-09-05-astra-vs-fable-51-five-builds/ https://openclawdatabase.com/news/videos/2026-09-05-astra-vs-fable-51-five-builds/ Sat, 05 Sep 2026 12:00:00 GMT The same five prompts given to GPT-6 Astra and Claude Fable 5.1 in parallel — a full-stack CRM as a native Mac app, a Notion clone, a clothing brand storefront, a 3D city model, and an Age of Empires II replica. Every session produced a working app with no steering, and the differences fall into a consistent pattern. https://openclawdatabase.com/news/videos/2026-09-03-agent-anatomy-harness-mcp-sandbox/ https://openclawdatabase.com/news/videos/2026-09-03-agent-anatomy-harness-mcp-sandbox/ Thu, 03 Sep 2026 12:00:00 GMT A plain-English breakdown of what an agent is made of beyond the model — the harness that runs the loop, MCP as the tool protocol, skills as loaded-on-demand instructions, the sandbox for code execution, and the production layer of subagents, approvals and observability — followed by a live build using an open-source harness. https://openclawdatabase.com/news/videos/2026-09-03-grok-bot-daily-briefing-podcast/ https://openclawdatabase.com/news/videos/2026-09-03-grok-bot-daily-briefing-podcast/ Thu, 03 Sep 2026 12:00:00 GMT A reproducible build: a personal chief-of-staff agent assembles a daily briefing from calendar, email and news, generates it as audio through an MCP-connected voice service, and pushes it to a private podcast feed so it arrives in a normal podcast app every morning. About twenty minutes of work, mostly conversation. https://openclawdatabase.com/news/videos/2026-09-02-agentic-engineering-setup-6775-sessions/ https://openclawdatabase.com/news/videos/2026-09-02-agentic-engineering-setup-6775-sessions/ Wed, 02 Sep 2026 12:00:00 GMT A full walkthrough of one practitioner's agentic engineering stack: the interfaces used to run many agents at once, how to combine subscriptions instead of paying API rates, and a step-by-step build of self-hosted cloud agents on a VPS driven entirely by an agent in plain English. https://openclawdatabase.com/news/videos/2026-09-02-anthropic-fable-51-prompting-techniques/ https://openclawdatabase.com/news/videos/2026-09-02-anthropic-fable-51-prompting-techniques/ Wed, 02 Sep 2026 12:00:00 GMT Four prompting techniques pulled from Anthropic's own Fable 5.1 documentation, with the supporting quotes: describe the finished state rather than the task list, match effort level to the work instead of leaving it on high, make the model verify its own output before handing it to you, and delegate to parallel subagents so the flagship drives strategy rather than doing the work. https://openclawdatabase.com/news/videos/2026-09-01-openclaw-2-clean-install-tour/ https://openclawdatabase.com/news/videos/2026-09-01-openclaw-2-clean-install-tour/ Tue, 01 Sep 2026 12:00:00 GMT A first-run walkthrough of OpenClaw 2.0 covering the shortened install, the automatic scan that finds local models and existing subscriptions, the three-stage memory process, the Lab section of pre-release toggles, and a complete Telegram channel pairing. https://openclawdatabase.com/news/videos/2026-08-31-grok-bot-agent-team-setup/ https://openclawdatabase.com/news/videos/2026-08-31-grok-bot-agent-team-setup/ Mon, 31 Aug 2026 12:00:00 GMT A full setup walkthrough for Grok Bot, xAI's always-on agent product, from a fresh account to a working multi-agent team. Covers the fields that actually control routing between agents, the shared-computer model behind each bot's own screen, scheduled routines with their trigger options, and an org topology that avoids one overloaded mega-agent. https://openclawdatabase.com/news/videos/2026-08-31-openclaw-2-upgrade-breakage-review/ https://openclawdatabase.com/news/videos/2026-08-31-openclaw-2-upgrade-breakage-review/ Mon, 31 Aug 2026 12:00:00 GMT A critical review of OpenClaw 2.0 from someone who left the project four months ago over update reliability. Covers the eight headline features, the in-place upgrade failing again and forcing a clean install, and the significant limitation that most of the new functionality exists only in the web app. https://openclawdatabase.com/news/videos/2026-08-31-rhyme-cli-agent-voice-replies/ https://openclawdatabase.com/news/videos/2026-08-31-rhyme-cli-agent-voice-replies/ Mon, 31 Aug 2026 12:00:00 GMT A small, portable trick: install a text-to-speech CLI, then instruct your coding agent to invoke it when a task finishes so it speaks a one-or-two sentence summary and a next-step question instead of adding to a wall of text. Works across harnesses because the agent is just running a command. https://openclawdatabase.com/news/videos/2026-08-24-fusion-harness-multi-model-debate/ https://openclawdatabase.com/news/videos/2026-08-24-fusion-harness-multi-model-debate/ Mon, 24 Aug 2026 12:00:00 GMT IndyDevDan walks through three multi-agent orchestration commands — opinion, debate, and collaborate — that run several models against one prompt, with per-run cost figures and the model-aliasing gotcha that stops agents sabotaging each other. https://openclawdatabase.com/news/videos/2026-08-24-sie-one-server-agent-models/ https://openclawdatabase.com/news/videos/2026-08-24-sie-one-server-agent-models/ Mon, 24 Aug 2026 12:00:00 GMT Fahd Mirza installs the Superlinked Inference Engine on a local GPU box and runs embeddings, re-ranking, entity extraction, and text generation through one client — encode, score, extract, generate — with models loaded on demand. https://openclawdatabase.com/news/videos/2026-08-23-codex-desktop-business-setup/ https://openclawdatabase.com/news/videos/2026-08-23-codex-desktop-business-setup/ Sun, 23 Aug 2026 12:00:00 GMT Nick Saraev's free Codex course walks the desktop app end to end — install and sign-in, importing your Claude Code setup, approval modes, the model and effort pickers, scheduled autonomous tasks, plugins and skill recording. https://openclawdatabase.com/news/videos/2026-08-23-deepseek-harness-vs-claude-code/ https://openclawdatabase.com/news/videos/2026-08-23-deepseek-harness-vs-claude-code/ Sun, 23 Aug 2026 12:00:00 GMT Nate Herk runs DeepSeek's open-source harness against Claude Code on identical prompts and the same model — 50 seconds vs 5 minutes on search, 3 minutes vs 17 on a deliverable — plus its four modes, its context injection, and the preview bugs. https://openclawdatabase.com/news/videos/2026-08-22-agent-budgets-not-tokens/ https://openclawdatabase.com/news/videos/2026-08-22-agent-budgets-not-tokens/ Sat, 22 Aug 2026 12:00:00 GMT An Anthropic CI engineer replaces scoped tokens with four enforceable primitives — asymmetric verbs, refilling rate limits, trip wires over allow lists, and the undo test — after an agent deleted 200 workloads in 90 seconds. https://openclawdatabase.com/news/videos/2026-08-21-claude-code-non-code-workspace-scaffold/ https://openclawdatabase.com/news/videos/2026-08-21-claude-code-non-code-workspace-scaffold/ Fri, 21 Aug 2026 12:00:00 GMT Nate Herk builds a Claude Code project for work that involves no code at all — a CLAUDE.md that indexes a context folder, assets and output directories, an interview skill for extracting context from your head, and the click path for wiring a custom MCP connector. https://openclawdatabase.com/news/videos/2026-08-21-free-models-claude-code-openrouter/ https://openclawdatabase.com/news/videos/2026-08-21-free-models-claude-code-openrouter/ Fri, 21 Aug 2026 12:00:00 GMT Point Claude Code at OpenRouter's free models by editing the env block in your settings file. Skills, CLAUDE.md and sub-agents all still fire — but jobs that normally take 20 minutes took six hours. https://openclawdatabase.com/news/videos/2026-08-21-uber-agentic-sdlc-software-factory/ https://openclawdatabase.com/news/videos/2026-08-21-uber-agentic-sdlc-software-factory/ Fri, 21 Aug 2026 12:00:00 GMT Uber's engineering leads walk through the six platform pieces behind their "software factory" — a model gateway, an MCP gateway with four generations of token optimisation, pre-provisioned agent dev pods, a 2,500-skill marketplace, a 40M-node context graph, and the assistant that ties them together. https://openclawdatabase.com/news/videos/2026-08-20-context-as-code-content-machine/ https://openclawdatabase.com/news/videos/2026-08-20-context-as-code-content-machine/ Thu, 20 Aug 2026 12:00:00 GMT Alex Lieberman and Dan of 10X break down a content pipeline built entirely from skills and markdown — a process layer in git, a personal voice layer on disk, six interviewer personas as six skills, an editor council with a 9-out-of-10 gate, and a lessons file the agent re-reads before every write. https://openclawdatabase.com/news/videos/2026-08-20-cto-overnight-agent-loop/ https://openclawdatabase.com/news/videos/2026-08-20-cto-overnight-agent-loop/ Thu, 20 Aug 2026 12:00:00 GMT Hursh Agrawal, CTO of The Browser Company, breaks down the 5pm-to-morning agent loop he uses to ship code around a manager's calendar — context gathering, verification baked into the prompt, and the org scaffolding that makes it safe. https://openclawdatabase.com/news/videos/2026-08-20-docker-agent-sandbox-runtime/ https://openclawdatabase.com/news/videos/2026-08-20-docker-agent-sandbox-runtime/ Thu, 20 Aug 2026 12:00:00 GMT Docker's Tushar Jain demos a micro-VM runtime that runs any agent harness inside a sandbox with injected credential stubs and a network policy — and argues you should split one task across two narrowly-scoped sandboxes instead of handing one agent every credential it might need. https://openclawdatabase.com/news/videos/2026-08-20-free-openrouter-model-opencode/ https://openclawdatabase.com/news/videos/2026-08-20-free-openrouter-model-opencode/ Thu, 20 Aug 2026 12:00:00 GMT A practical heuristic for choosing among OpenRouter's free models — sort by output price, then rank by cumulative tokens used — plus the OpenCode /skills and /models flow used to run a real competitor-schema audit on a zero-cost stack. https://openclawdatabase.com/news/videos/2026-08-20-hugging-face-agent-paper-outreach/ https://openclawdatabase.com/news/videos/2026-08-20-hugging-face-agent-paper-outreach/ Thu, 20 Aug 2026 12:00:00 GMT Niels Rogge shows the two-generation architecture behind an agent that opens thousands of GitHub issues a night — a deterministic GitHub Actions cron pipeline first, then a Claude Agent SDK agent on Modal batch containers with exactly one CLI, one skill and a sandbox. https://openclawdatabase.com/news/videos/2026-08-20-zed-delta-agent-thread-workflow/ https://openclawdatabase.com/news/videos/2026-08-20-zed-delta-agent-thread-workflow/ Thu, 20 Aug 2026 12:00:00 GMT A hands-on walkthrough of Zed's new standalone agent app, Delta — the Delta DB version-control layer that links code to the conversation that produced it, agent-isolated worktree clones, bring-your-own API keys, and Claude Code sync over ACP. https://openclawdatabase.com/news/videos/2026-08-19-berd-open-source-agent-desktop/ https://openclawdatabase.com/news/videos/2026-08-19-berd-open-source-agent-desktop/ Wed, 19 Aug 2026 12:00:00 GMT Block open-sourced Berd, an Apache-2 desktop app that fronts Claude Code, Codex, Copilot, AMP, Cursor Agent and Goose through the Agent Client Protocol — with a model and reasoning-effort picker, a skills menu, and one install failure worth knowing about. https://openclawdatabase.com/news/videos/2026-08-19-grok-bot-chief-of-staff-skills/ https://openclawdatabase.com/news/videos/2026-08-19-grok-bot-chief-of-staff-skills/ Wed, 19 Aug 2026 12:00:00 GMT A power-user pass over Grok Bot — the verbatim chief-of-staff delegation description, the split between per-bot and shared memory, bridging missing integrations through Composio, logging agent work to ClickUp, and routines with event triggers. https://openclawdatabase.com/news/videos/2026-08-19-schema-grounded-agent-harness/ https://openclawdatabase.com/news/videos/2026-08-19-schema-grounded-agent-harness/ Wed, 19 Aug 2026 12:00:00 GMT A healthcare-claims talk with a transferable lesson for any agent builder: constrain the agent's output to an existing public standard instead of a model-invented schema, move memory out of local files, and treat every external system as untrustworthy by default. https://openclawdatabase.com/news/videos/2026-08-18-hermes-loop-recurring-agent-tasks/ https://openclawdatabase.com/news/videos/2026-08-18-hermes-loop-recurring-agent-tasks/ Tue, 18 Aug 2026 12:00:00 GMT Hermes shipped /loop — a timer-driven wake-up that re-reads the world each tick and stops on a condition you set. Full walkthrough of the loop settings, a working deploy-watcher demo, and when to use /loop vs /goal vs cron. https://openclawdatabase.com/news/videos/2026-08-17-claw-patrol-agent-firewall-deno/ https://openclawdatabase.com/news/videos/2026-08-17-claw-patrol-agent-firewall-deno/ Mon, 17 Aug 2026 12:00:00 GMT Ryan Dahl on giving OpenClaw read-write access to production Postgres, Kubernetes, ClickHouse and AWS — and the MIT-licensed proxy Deno built so that is survivable. Parses every byte below HTTP, understands the Postgres wire protocol, injects credentials the agent never sees, and routes approvals to Slack or an LLM judge. https://openclawdatabase.com/news/videos/2026-08-17-grok-bot-setup-agent-mail-plugins/ https://openclawdatabase.com/news/videos/2026-08-17-grok-bot-setup-agent-mail-plugins/ Mon, 17 Aug 2026 12:00:00 GMT A first-week Grok Bot setup pass — the brain-dump then reverse-prompt method for configuring bots, which plugins to install, and the account-isolation pattern of giving every agent its own email address so you never hand it your own login. https://openclawdatabase.com/news/videos/2026-08-14-codex-vs-claude-code-same-app/ https://openclawdatabase.com/news/videos/2026-08-14-codex-vs-claude-code-same-app/ Fri, 14 Aug 2026 12:00:00 GMT One prompt, two coding agents, two finished apps. Measured side by side: hours, dollars, output tokens, sub-agent counts, tool calls and test counts — plus the prompting rule that explains why the cheaper run produced the better product.